Embedders

Convert text into vector representations for semantic search.

Embedders convert text into vectors (lists of numbers) that capture meaning. These vectors enable semantic search, so “How do I reset my passcode?” finds documents mentioning “change PIN” even without keyword matches.

from agno.knowledge.knowledge import Knowledge
from agno.knowledge.embedder.openai import OpenAIEmbedder
from agno.vectordb.pgvector import PgVector

knowledge = Knowledge(
    vector_db=PgVector(
        table_name="docs",
        db_url=db_url,
        embedder=OpenAIEmbedder(),  # Default
    ),
)

How It Works

1. Insert: When you add content, each chunk is converted to a vector
2. Store: Vectors are saved in your vector database
3. Search: Queries are embedded and matched against stored vectors by similarity

Many vector database implementations, including PgVector, use OpenAIEmbedder when embedder is omitted. Provider-hosted implementations can differ. UpstashVectorDb uses Upstash hosted embeddings when embedder is omitted.

Configuration

from agno.knowledge.embedder.openai import OpenAIEmbedder

embedder = OpenAIEmbedder(
    id="text-embedding-3-small",
    dimensions=1536,
)

Using with Knowledge

from agno.knowledge.knowledge import Knowledge
from agno.vectordb.pgvector import PgVector

knowledge = Knowledge(
    vector_db=PgVector(
        table_name="docs",
        db_url=db_url,
        embedder=OpenAIEmbedder(id="text-embedding-3-small"),
    ),
)

# Content is embedded automatically on insert
knowledge.insert(path="documents/")

Batch Embeddings

Process multiple texts in a single API call to reduce requests and improve performance:

embedder = OpenAIEmbedder(
    id="text-embedding-3-small",
    dimensions=1536,
    enable_batch=True,
    batch_size=100,
)

Embedders with batch support: OpenAI, Azure OpenAI, Gemini, Cohere, Voyage AI, Mistral, Fireworks, Together, Jina, Nebius, LangDB, vLLM.

Best Practices

Re-embed when changing models: Vectors from different embedders aren’t compatible. If you switch embedders, you must re-embed all content.

Test retrieval quality: Use sample queries to verify you’re finding the right chunks. Adjust chunking strategy or embedder if results are poor.

Match dimensions: Ensure your embedder’s output dimensions match what your vector database expects.

Supported Embedders

Embedder	Type	Cost	Notes
OpenAI	Hosted	$$	Default for PgVector and many other vector databases
Gemini	Hosted	$$	Multilingual, Google ecosystem
Cohere	Hosted	$$	Strong retrieval performance
Voyage AI	Hosted	$$$	Specialized for retrieval
Mistral	Hosted	$$	European provider
Ollama	Local	Free	Privacy, offline
FastEmbed	Local	Free	Fast local embeddings
SentenceTransformers	Local	Free	Local models via sentence-transformers
vLLM	Local	Free	Local model or self-hosted vLLM server
HuggingFace	Hosted	Free/$	Open source models via Inference API
AWS Bedrock	Hosted	$$	AWS ecosystem
Azure OpenAI	Hosted	$$	Azure ecosystem
Fireworks	Hosted	$	Fast inference
Together	Hosted	$	Open source models
Jina	Hosted	$$	Multilingual
Nebius	Hosted	$	European provider
LangDB	Hosted	$$	AI gateway, OpenAI-compatible

Choosing an Embedder

Consideration	Recommendation
General use	OpenAI or Gemini
Privacy/offline	Ollama or FastEmbed
Multilingual	Gemini or Jina
Cost-sensitive	Local embedders (free) or Fireworks/Together ($)
Best retrieval quality	Voyage AI or Cohere

Key factors:

• Hosted vs local: Local for privacy and no API costs; hosted for quality and convenience
• Latency and cost: Smaller models are cheaper and faster; larger models often retrieve better
• Language support: Ensure your embedder supports your content’s languages
• Dimension size: Match your vector database’s expected embedding dimensions

Next Steps

OpenAI Embedder

Default embedder setup

Ollama Embedder

Local embeddings for privacy

Vector DB

Store your embeddings

Chunking

Prepare content for embedding